Unix/Linux Go Back    


War Stories Tell your work related tech stories and share experiences here. Share your successes and failures and other "war stories" in this forum.

Linus Torvalds reply about Meltdown and Spectre.

War Stories


Reply    
 
Thread Tools Search this Thread Display Modes
    #1  
Old Unix and Linux 3 Weeks Ago   -   Original Discussion by wisecracker
wisecracker's Unix or Linux Image
wisecracker wisecracker is offline
Registered User
 
Join Date: Jan 2013
Last Activity: 6 February 2018, 3:15 PM EST
Location: Loughborough
Posts: 1,247
Thanks: 373
Thanked 331 Times in 258 Posts
Linus Torvalds reply about Meltdown and Spectre.

Apologies if this is the wrong forum but...

...This is hard hitting stuff.

LKML: Linus Torvalds: Re: [RFC 09/10] x86/enter: Create macros to restrict/unrestrict Indirect Branch Speculation
The Following 4 Users Say Thank You to wisecracker For This Useful Post:
Corona688 (3 Weeks Ago), dodona (3 Weeks Ago), hicksd8 (3 Weeks Ago), rbatte1 (3 Weeks Ago)
Sponsored Links
    #2  
Old Unix and Linux 3 Weeks Ago   -   Original Discussion by wisecracker
dodona's Unix or Linux Image
dodona dodona is offline
Registered User
 
Join Date: Jun 2012
Last Activity: 19 February 2018, 9:07 AM EST
Posts: 25
Thanks: 66
Thanked 5 Times in 4 Posts
Quote:
As it is, the patches are COMPLETE AND UTTER GARBAGE.
.. for a only hypothetical problem.
Sponsored Links
    #3  
Old Unix and Linux 3 Weeks Ago   -   Original Discussion by wisecracker
wisecracker's Unix or Linux Image
wisecracker wisecracker is offline
Registered User
 
Join Date: Jan 2013
Last Activity: 6 February 2018, 3:15 PM EST
Location: Loughborough
Posts: 1,247
Thanks: 373
Thanked 331 Times in 258 Posts
Quote:
Originally Posted by dodona View Post
.. for a only hypothetical problem.
Perhaps, perhaps not; a hardware design problem that goes back decades just to get performance figures to sell their CPUs is more than bad to say the least - especially when the manufacturers knew about it.
Also, it can't be that hypothetical if it has been proven to work. It is the big/gigantic sytems that will be affected by these patches not so much the piffling little stuff like my MacBook Pro. I can handle a performance hit knowing about it now but can the big guns? <- Rhetorical!

But unless one has experience of writing kernels and OSes then one can't possibly know how difficult it is to create patches, (that will inherently give a performance hit just to _correct_ a deliberate manufacturing fault for profit), that will not slow things down too much.
    #4  
Old Unix and Linux 3 Weeks Ago   -   Original Discussion by wisecracker
MadeInGermany's Unix or Linux Image
MadeInGermany MadeInGermany is offline Forum Staff  
Moderator
 
Join Date: May 2012
Last Activity: 20 February 2018, 2:37 AM EST
Location: Simplicity
Posts: 3,981
Thanks: 342
Thanked 1,337 Times in 1,206 Posts
The x86 monoculture leads to risks like this.
See also --> OS monoculture...
Sponsored Links
    #5  
Old Unix and Linux 2 Weeks Ago   -   Original Discussion by wisecracker
dodona's Unix or Linux Image
dodona dodona is offline
Registered User
 
Join Date: Jun 2012
Last Activity: 19 February 2018, 9:07 AM EST
Posts: 25
Thanks: 66
Thanked 5 Times in 4 Posts
I read that with the 4.15 Kernel with build-in patches for the hypothetical problem comes with 15-30% performance loss. Immediately I thought 'wow, Intel, AMD and the hardware pushers will earn a lot of $'. I mean todays cpu is so fast that there isn't a need for a upgrade. However the hypothetical problem with the 15-30% performance loss is the need for an upgradeLinux. Again everything comes down to $, and nothing else than $.Linux
Sponsored Links
    #6  
Old Unix and Linux 2 Weeks Ago   -   Original Discussion by wisecracker
Corona688's Unix or Linux Image
Corona688 Corona688 is offline Forum Staff  
Mead Rotor
 
Join Date: Aug 2005
Last Activity: 16 February 2018, 12:07 PM EST
Location: Saskatchewan
Posts: 22,613
Thanks: 1,169
Thanked 4,301 Times in 3,968 Posts
Quote:
Originally Posted by dodona View Post
I read that with the 4.15 Kernel with build-in patches for the hypothetical problem comes with 15-30% performance loss.
That's way overblown, because the performance loss happens during the switch between userspace and kernelspace. Sensible programs don't spend most of their time doing thousands of tiny system calls.

A 30% reduction in performance there is bad, but not the same as a 30% slower computer.
Sponsored Links
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Linux More UNIX and Linux Forum Topics You Might Find Helpful
Thread Thread Starter Forum Replies Last Post
AIX 6.1, POWER5 and Spectre/Meltdown the_garbage AIX 7 2 Weeks Ago 11:41 AM
Meltdown and Spectre CPU bugs Scrutinizer What is on Your Mind? 6 4 Weeks Ago 03:27 PM
A Couple of Short Linus Torvalds Videos Perderabo News, Links, Events and Announcements 0 01-23-2007 01:28 AM



All times are GMT -4. The time now is 05:47 AM.